Echo-ey AC3, buffer too small?

[Eric Lund]

Folks,

I am, for the first time, trying to play ATSC streams with AC3
audio in them on my MVP.  I have had fairly limited luck, and am
planning to move to a Popcorn Hour box shortly for reliable
playback, but I seem to be tantalizingly close to being able to
play most of the streams.   The major problem I am seeing is with
the AC3 audio, which, when it plays back at all, plays back with
several distinct echoes (it sounds like the individual sub-streams
are being mixed down at about 1/4 second intervals).

has anyone heard this?  I am looking around for a cause, and I don't
see any obvious one unless it is running out of buffer space.

                                Eric

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[MVallevand]

On Tue, Feb 10, 2009 at 6:01 PM, Eric Lund <eric@...> wrote:

> play most of the streams.   The major problem I am seeing is with
> the AC3 audio, which, when it plays back at all, plays back with
> several distinct echoes (it sounds like the individual sub-streams
> are being mixed down at about 1/4 second intervals).
>
> has anyone heard this?  I am looking around for a cause, and I don't
> see any obvious one unless it is running out of buffer space.

I did hear this when I first installed my HVR 1600 and I didn't have a
good stream so many bogus audio tracks were showing when I checked
through the menu key.   Getting a reliable ts stream solved this to
the point where I could complain to myself about the sync and wish I
understood how to control the sync in the demux like Simon does. I do
get occasional drop outs with the software decoder but in digital ac3
pass thru this isn't too much of a problem.

Martin

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

Martin,

I only see one AC3 stream when I am hearing this (looking in
the menu).  I took another look and I was way overestimating the
time between the "echoes".  I can't count them easily, but there
are at least 3 (an educated guess would be 5, but that could be
misleading too) distinct sounds about 0.1 seconds apart.  It seems
like I am getting a good stream just not decoding it quite right.
There is an XXX comment in a52_decode_data() (audio.c) that refers
to reworking the AC3 decoding to handle partial buffers.  Since the
new ATSC streams seem to be structured differently (and a bit higher
bit rate than my "analog" recordings typically were) I wonder whether
that is what I am running into.

The demuxer / decoder code is one part where I have spent little or no
time.  That was originally Jon's stuff.  So, now I am digging around in
there and I am finding it less obvious that I had thought it might be
:).

                                Eric

MVallevand wrote:

> On Tue, Feb 10, 2009 at 6:01 PM, Eric Lund <eric@...> wrote:
>
>> play most of the streams.   The major problem I am seeing is with
>> the AC3 audio, which, when it plays back at all, plays back with
>> several distinct echoes (it sounds like the individual sub-streams
>> are being mixed down at about 1/4 second intervals).
>>
>> has anyone heard this?  I am looking around for a cause, and I don't
>> see any obvious one unless it is running out of buffer space.
>
> I did hear this when I first installed my HVR 1600 and I didn't have a
> good stream so many bogus audio tracks were showing when I checked
> through the menu key.   Getting a reliable ts stream solved this to
> the point where I could complain to myself about the sync and wish I
> understood how to control the sync in the demux like Simon does. I do
> get occasional drop outs with the software decoder but in digital ac3
> pass thru this isn't too much of a problem.
>
> Martin
>
> ------------------------------------------------------------------------------
> Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
> software. With Adobe AIR, Ajax developers can use existing skills and code to
> build responsive, highly engaging applications that combine the power of local
> resources and data with the reach of the web. Download the Adobe AIR SDK and
> Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
> _______________________________________________
> Mvpmc-devel mailing list
> Mvpmc-devel@...
> https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
>

------------------------------------------------------------------------------
Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM)
software. With Adobe AIR, Ajax developers can use existing skills and code to
build responsive, highly engaging applications that combine the power of local
resources and data with the reach of the web. Download the Adobe AIR SDK and
Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[MVallevand]

On Wed, Feb 11, 2009 at 8:58 AM, Eric Lund <eric@...> wrote:

> I only see one AC3 stream when I am hearing this (looking in
> the menu).  I took another look and I was way overestimating the
> time between the "echoes".  I can't count them easily, but there

This is different then I hear.  Typically I just see the odd "error"
message from line  1115 with audio breakup.   Does the audio stream
decode okay by itself?   You should be able to extract it  like this.

mplayer -dumpaudio -dumpfile audio.ac3  video.ts

Martin.

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

Martin,

The audio stream extracted as you describe plays cleanly on the MVP.
The entire ATSC stream plays cleanly under mplayer on my laptop.  I
can't get my host mode build to play anything (first experiment with
host mode, probably entirely my fault), so I have not yet been able to
experiment with it under mvpmc on my laptop (I thought this would be a
good experiment, I will keep trying).

When I run Roger's makempeg2 transcoder against it, I get a handful of
duplicate frames initially, and the following:

    Try adding the scale filter, e.g. -vf spp,scale instead of -vf spp.
    Limiting audio preload to 0.4s.
    Increasing audio density to 4.

The echoes go away when transcoded (but the audio stream is now mpeg not
ac3).  The sync is screwed up (as probably should be expected).  I have
not always had that luck with my ATSC recordings, but the symptoms are
different on the ones that won't transcode cleanly.  The information
from the transcoder script suggest to me that the audio in the original
stream is too sparsely packed and the video is overrunning it.

I don't know enough about mencoder to know for sure what options
to put in to convince it to pack the AC3 more densely without
transcoding it to mpeg.  Here is the mencoder command that makempeg2
composed for the stream in question:

     mencoder -of mpeg -mpegopts format=dvd  -ovc copy -srate 48000  \
              -oac mp3lame -lameopts cbr:br=192:preset=192 \
              -audio-preload 0.0 3021_20090205210000.mpg \
              -o 3021_20090205210000.mpg.mpg

The mencoder manual page is a bit imtimidating [this from a long
time kernel developer] :).

                                 Eric



MVallevand wrote:
 > On Wed, Feb 11, 2009 at 8:58 AM, Eric Lund <eric@...>
wrote:
 >
 >> I only see one AC3 stream when I am hearing this (looking in
 >> the menu).  I took another look and I was way overestimating the
 >> time between the "echoes".  I can't count them easily, but there
 >
 > This is different then I hear.  Typically I just see the odd "error"
 > message from line  1115 with audio breakup.   Does the audio stream
 > decode okay by itself?   You should be able to extract it  like this.
 >
 > mplayer -dumpaudio -dumpfile audio.ac3  video.ts
 >
 > Martin.
 >
 >
------------------------------------------------------------------------------
 > Open Source Business Conference (OSBC), March 24-25, 2009, San
Francisco, CA
 > -OSBC tackles the biggest issue in open source: Open Sourcing the
Enterprise
 > -Strategies to boost innovation and cut costs with open source
participation
 > -Receive a $600 discount off the registration fee with the source
code: SFAD
 > http://p.sf.net/sfu/XcvMzF8H
 > _______________________________________________
 > Mvpmc-devel mailing list
 > Mvpmc-devel@...
 > https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
 >

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

Martin,

I managed to get the transcoded files to play cleanly with the sync
pretty darn close (close enough to fix with a few record button presses)
by changing the script to generate the following:

     mencoder -of mpeg -mpegopts format=mpeg  -audio-density 4
              -ovc copy -srate 48000 -oac mp3lame
              -lameopts cbr:br=192:preset=192 \
              -audio-preload 0.0 3021_20090205210000.mpg \
              -o 3021_20090205210000.mpg.mpg

the difference being format=mpeg and -audio-density 4 being explicit.

This seems to work on my setup.  I will have to rebuild my dongle.bin
with a standard sized demuxer buffer, though, to be sure.

I would prefer not having to transcode, but, since I will probably have
to anyway for the HD stuff that I will soon be seeing, I guess I can
live with it.

As the brave new world unfurls I will probably have other problems.

                               Eric

Eric Lund wrote:
 > Martin,
 >
 > The audio stream extracted as you describe plays cleanly on the MVP.
 > The entire ATSC stream plays cleanly under mplayer on my laptop.  I
 > can't get my host mode build to play anything (first experiment with
 > host mode, probably entirely my fault), so I have not yet been able to
 > experiment with it under mvpmc on my laptop (I thought this would be a
 > good experiment, I will keep trying).
 >
 > When I run Roger's makempeg2 transcoder against it, I get a handful of
 > duplicate frames initially, and the following:
 >
 >     Try adding the scale filter, e.g. -vf spp,scale instead of -vf spp.
 >     Limiting audio preload to 0.4s.
 >     Increasing audio density to 4.
 >
 > The echoes go away when transcoded (but the audio stream is now mpeg not
 > ac3).  The sync is screwed up (as probably should be expected).  I have
 > not always had that luck with my ATSC recordings, but the symptoms are
 > different on the ones that won't transcode cleanly.  The information
 > from the transcoder script suggest to me that the audio in the original
 > stream is too sparsely packed and the video is overrunning it.
 >
 > I don't know enough about mencoder to know for sure what options
 > to put in to convince it to pack the AC3 more densely without
 > transcoding it to mpeg.  Here is the mencoder command that makempeg2
 > composed for the stream in question:
 >
 >      mencoder -of mpeg -mpegopts format=dvd  -ovc copy -srate 48000  \
 >               -oac mp3lame -lameopts cbr:br=192:preset=192 \
 >               -audio-preload 0.0 3021_20090205210000.mpg \
 >               -o 3021_20090205210000.mpg.mpg
 >
 > The mencoder manual page is a bit imtimidating [this from a long
 > time kernel developer] :).
 >
 >                                  Eric
 >
 >
 >
 > MVallevand wrote:
 >  > On Wed, Feb 11, 2009 at 8:58 AM, Eric Lund <eric@...>
 > wrote:
 >  >
 >  >> I only see one AC3 stream when I am hearing this (looking in
 >  >> the menu).  I took another look and I was way overestimating the
 >  >> time between the "echoes".  I can't count them easily, but there
 >  >
 >  > This is different then I hear.  Typically I just see the odd "error"
 >  > message from line  1115 with audio breakup.   Does the audio stream
 >  > decode okay by itself?   You should be able to extract it  like this.
 >  >
 >  > mplayer -dumpaudio -dumpfile audio.ac3  video.ts
 >  >
 >  > Martin.
 >  >
 >  >
 > ------------------------------------------------------------------------------
 >  > Open Source Business Conference (OSBC), March 24-25, 2009, San
 > Francisco, CA
 >  > -OSBC tackles the biggest issue in open source: Open Sourcing the
 > Enterprise
 >  > -Strategies to boost innovation and cut costs with open source
 > participation
 >  > -Receive a $600 discount off the registration fee with the source
 > code: SFAD
 >  > http://p.sf.net/sfu/XcvMzF8H
 >  > _______________________________________________
 >  > Mvpmc-devel mailing list
 >  > Mvpmc-devel@...
 >  > https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
 >  >
 >
 > ------------------------------------------------------------------------------
 > Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
 > -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
 > -Strategies to boost innovation and cut costs with open source participation
 > -Receive a $600 discount off the registration fee with the source code: SFAD
 > http://p.sf.net/sfu/XcvMzF8H
 > _______________________________________________
 > Mvpmc-devel mailing list
 > Mvpmc-devel@...
 > https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
 >

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Simon Hyde]

On Tue, 17 Feb 2009, Eric Lund wrote:

> Martin,
>
> I managed to get the transcoded files to play cleanly with the sync
> pretty darn close (close enough to fix with a few record button presses)
> by changing the script to generate the following:
>
>     mencoder -of mpeg -mpegopts format=mpeg  -audio-density 4
>              -ovc copy -srate 48000 -oac mp3lame
>              -lameopts cbr:br=192:preset=192 \
>              -audio-preload 0.0 3021_20090205210000.mpg \
>              -o 3021_20090205210000.mpg.mpg

FWIW this all seems a little over-complicated (as does Roger Heflin's
transcode script). I find the command line required for ffmpeg to be much
simpler than for mencoder.

To transcode everything (useful for non-MPEG1/2 encoded files):

ffmpeg -i source.avi -target pal-dvd -acodec libmp3lame output.mpg
or
ffmpeg -i source.avi -target pal-svcd output.mpg
or
ffmpeg -i source.avi -target pal-vcd output.mpg

Take your pick from the above depending on the quality/bitrate of MPEG2


If your file already contains MPEG1/2 video of an appropriate resolution
then you can just go for:

ffmpeg -i source.mpg -target pal-svcd -vcodec copy output.mpg

Obviously for NTSC-resolution video you'll have to change the pal-dvd,
pal-svcd or pal-vcd to ntsc-dvd/ntsc-svcd/ntsc-vcd

If you really want widescreen programs that come out of the end of this to
be letterboxed so that you can leave your TV permanently in zoom mode
(which I believe Roger Heflin's scripts produce), then just put your MVP
into 4:3 mode rather than 16:9 mode.

Cheers,

Simon

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[stuart-20]

Simon Hyde wrote:

> On Tue, 17 Feb 2009, Eric Lund wrote:
>
>> Martin,
>>
>> I managed to get the transcoded files to play cleanly with the sync
>> pretty darn close (close enough to fix with a few record button presses)
>> by changing the script to generate the following:
>>
>>     mencoder -of mpeg -mpegopts format=mpeg  -audio-density 4
>>              -ovc copy -srate 48000 -oac mp3lame
>>              -lameopts cbr:br=192:preset=192 \
>>              -audio-preload 0.0 3021_20090205210000.mpg \
>>              -o 3021_20090205210000.mpg.mpg
>
> FWIW this all seems a little over-complicated (as does Roger Heflin's
> transcode script). I find the command line required for ffmpeg to be much
> simpler than for mencoder.
>
> To transcode everything (useful for non-MPEG1/2 encoded files):
>
> ffmpeg -i source.avi -target pal-dvd -acodec libmp3lame output.mpg
> or
> ffmpeg -i source.avi -target pal-svcd output.mpg
> or
> ffmpeg -i source.avi -target pal-vcd output.mpg
>
> Take your pick from the above depending on the quality/bitrate of MPEG2
>
>
> If your file already contains MPEG1/2 video of an appropriate resolution
> then you can just go for:
>
> ffmpeg -i source.mpg -target pal-svcd -vcodec copy output.mpg
>
> Obviously for NTSC-resolution video you'll have to change the pal-dvd,
> pal-svcd or pal-vcd to ntsc-dvd/ntsc-svcd/ntsc-vcd
>
> If you really want widescreen programs that come out of the end of this to
> be letterboxed so that you can leave your TV permanently in zoom mode
> (which I believe Roger Heflin's scripts produce), then just put your MVP
> into 4:3 mode rather than 16:9 mode.
>
> Cheers,
>
> Simon

Your timing is impeccable Simon.  As I was just searching for a trancode
command suitable for Chase's recent lib changes for the mvpmc box.  So,
as ATSC recordings are already MPEG2 (I assume) would you recommend the
last command you listed above for testing?:

ffmpeg -i source.mpg -target pal-svcd -vcodec copy output.mpg

(Note, nothing has changed in the mvpmc.org git repository.  All the lib
changes are in Chase's private git repository.  Doing a make of his copy
of the mvpmc project does not work - yet.)



------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Simon Hyde]

On Wed, 18 Feb 2009, stuart wrote:

> Your timing is impeccable Simon.  As I was just searching for a trancode
> command suitable for Chase's recent lib changes for the mvpmc box.  So, as
> ATSC recordings are already MPEG2 (I assume) would you recommend the last
> command you listed above for testing?:
>
> ffmpeg -i source.mpg -target pal-svcd -vcodec copy output.mpg

Yep, that should work fine for SD ATSC recordings (HD recordings need to
be transcoded because their resolution is too high for the MVP). For
correctness I'd use ntsc-svcd rather than pal-svcd, but I don't think
it'll actually make any difference since you're copying the video stream
anyway.

> (Note, nothing has changed in the mvpmc.org git repository.  All the lib
> changes are in Chase's private git repository.  Doing a make of his copy of
> the mvpmc project does not work - yet.)

No. I think the problem you're going to suffer from here is that his code
relies on SSL libraries which aren't part of the mvpmc build, so you'll
have to get a working SSL library linked in first.

Cheers,

Simon

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Chase Douglas-4]

On Feb 18, 2009, at 1:30 PM, Simon Hyde wrote:

>> (Note, nothing has changed in the mvpmc.org git repository.  All  
>> the lib
>> changes are in Chase's private git repository.  Doing a make of his  
>> copy of
>> the mvpmc project does not work - yet.)
>
> No. I think the problem you're going to suffer from here is that his  
> code
> relies on SSL libraries which aren't part of the mvpmc build, so  
> you'll
> have to get a working SSL library linked in first.

I should have mentioned this before. You'll need GnuTLS for the SSL/
TLS. I haven't modified the mvpmc makefiles to take this into account,  
but I assume it's as simple as adding -lgnutls to the linker flags.

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Simon Hyde]

On Wed, 18 Feb 2009, Chase Douglas wrote:

>> No. I think the problem you're going to suffer from here is that his code
>> relies on SSL libraries which aren't part of the mvpmc build, so you'll
>> have to get a working SSL library linked in first.
>
> I should have mentioned this before. You'll need GnuTLS for the SSL/TLS. I
> haven't modified the mvpmc makefiles to take this into account, but I assume
> it's as simple as adding -lgnutls to the linker flags.

Sadly not. The mvpmc build system also builds all the required libraries
that are installed on the embedded MVP device. At the moment i'm not aware
of anything on the mvpmc that requires GnuTLS, so someone will have to
setup the mvpmc build system to build the GnuTLS library aswell.

Cheers,

Simon

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Chase Douglas-4]

On Feb 18, 2009, at 1:44 PM, Simon Hyde wrote:

> On Wed, 18 Feb 2009, Chase Douglas wrote:
>
>>> No. I think the problem you're going to suffer from here is that  
>>> his code
>>> relies on SSL libraries which aren't part of the mvpmc build, so  
>>> you'll
>>> have to get a working SSL library linked in first.
>>
>> I should have mentioned this before. You'll need GnuTLS for the SSL/
>> TLS. I haven't modified the mvpmc makefiles to take this into  
>> account, but I assume it's as simple as adding -lgnutls to the  
>> linker flags.
>
> Sadly not. The mvpmc build system also builds all the required  
> libraries that are installed on the embedded MVP device. At the  
> moment i'm not aware of anything on the mvpmc that requires GnuTLS,  
> so someone will have to setup the mvpmc build system to build the  
> GnuTLS library aswell.

That's unfortunate. Just FYI, gnutls requires libgcrypt, which  
requires libgpg-error. However, all three libs are easily cross  
compiled in my experience. I was able to incorporate them into my  
iPhone project in Xcode in a matter of about an hour or two.

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

May I suggest ifdefs here?  I don't think we want to open this
can of worms across the board (especially on the MVP which is
a bit memory limited).

                                Eric

Chase Douglas wrote:

> On Feb 18, 2009, at 1:44 PM, Simon Hyde wrote:
>> On Wed, 18 Feb 2009, Chase Douglas wrote:
>>
>>>> No. I think the problem you're going to suffer from here is that  
>>>> his code
>>>> relies on SSL libraries which aren't part of the mvpmc build, so  
>>>> you'll
>>>> have to get a working SSL library linked in first.
>>> I should have mentioned this before. You'll need GnuTLS for the SSL/
>>> TLS. I haven't modified the mvpmc makefiles to take this into  
>>> account, but I assume it's as simple as adding -lgnutls to the  
>>> linker flags.
>> Sadly not. The mvpmc build system also builds all the required  
>> libraries that are installed on the embedded MVP device. At the  
>> moment i'm not aware of anything on the mvpmc that requires GnuTLS,  
>> so someone will have to setup the mvpmc build system to build the  
>> GnuTLS library aswell.
>
> That's unfortunate. Just FYI, gnutls requires libgcrypt, which  
> requires libgpg-error. However, all three libs are easily cross  
> compiled in my experience. I was able to incorporate them into my  
> iPhone project in Xcode in a matter of about an hour or two.
>
> ------------------------------------------------------------------------------
> Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
> -OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
> -Strategies to boost innovation and cut costs with open source participation
> -Receive a $600 discount off the registration fee with the source code: SFAD
> http://p.sf.net/sfu/XcvMzF8H
> _______________________________________________
> Mvpmc-devel mailing list
> Mvpmc-devel@...
> https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
>

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Chase Douglas-4]

That's a valid suggestion, and of course mvpmc isn't my project so I  
have no say in the final decision. I would however suggest that either  
all of my changes in relation to MTP are ifdef'd or none are.  
Basically, MTP requires TLS-SRP for any of the added functionality. I  
don't want anyone enabling MTP and using it in an unencrypted fashion,  
exposing the completely unsecure MythTV protocol to the greater  
Internet.

Thanks

On Feb 18, 2009, at 2:10 PM, Eric Lund wrote:

> May I suggest ifdefs here?  I don't think we want to open this
> can of worms across the board (especially on the MVP which is
> a bit memory limited).
>
> Chase Douglas wrote:
>> On Feb 18, 2009, at 1:44 PM, Simon Hyde wrote:
>>> On Wed, 18 Feb 2009, Chase Douglas wrote:
>>>
>>>>> No. I think the problem you're going to suffer from here is  
>>>>> that  his code
>>>>> relies on SSL libraries which aren't part of the mvpmc build,  
>>>>> so  you'll
>>>>> have to get a working SSL library linked in first.
>>>> I should have mentioned this before. You'll need GnuTLS for the  
>>>> SSL/ TLS. I haven't modified the mvpmc makefiles to take this  
>>>> into  account, but I assume it's as simple as adding -lgnutls to  
>>>> the  linker flags.
>>> Sadly not. The mvpmc build system also builds all the required  
>>> libraries that are installed on the embedded MVP device. At the  
>>> moment i'm not aware of anything on the mvpmc that requires  
>>> GnuTLS,  so someone will have to setup the mvpmc build system to  
>>> build the  GnuTLS library aswell.
>> That's unfortunate. Just FYI, gnutls requires libgcrypt, which  
>> requires libgpg-error. However, all three libs are easily cross  
>> compiled in my experience. I was able to incorporate them into my  
>> iPhone project in Xcode in a matter of about an hour or two.

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

That sounds like a very reasonable suggestion.

I will say that this is more a concern about encouraging risky behavior
than causing it, though.  It is not clear that anyone would expose
unencrypted mythtv protocol outside their firewall (regardless of MTP)
anyway.  If, for some completely inexplicable reason, someone wants to
play with MTP inside their firewall, it's their choice.

Nevertheless, I don't see any objection to ifdef'ing the complete set of
changes (once we know they can be built in a friendly environment).

                                Eric

Chase Douglas wrote:
 > That's a valid suggestion, and of course mvpmc isn't my project so I
 > have no say in the final decision. I would however suggest that either
 > all of my changes in relation to MTP are ifdef'd or none are. Basically,
 > MTP requires TLS-SRP for any of the added functionality. I don't want
 > anyone enabling MTP and using it in an unencrypted fashion, exposing the
 > completely unsecure MythTV protocol to the greater Internet.
 >
 > Thanks
 >
 > On Feb 18, 2009, at 2:10 PM, Eric Lund wrote:
 >> May I suggest ifdefs here?  I don't think we want to open this
 >> can of worms across the board (especially on the MVP which is
 >> a bit memory limited).
 >>
 >> Chase Douglas wrote:
 >>> On Feb 18, 2009, at 1:44 PM, Simon Hyde wrote:
 >>>> On Wed, 18 Feb 2009, Chase Douglas wrote:
 >>>>
 >>>>>> No. I think the problem you're going to suffer from here is that
 >>>>>> his code
 >>>>>> relies on SSL libraries which aren't part of the mvpmc build, so
 >>>>>> you'll
 >>>>>> have to get a working SSL library linked in first.
 >>>>> I should have mentioned this before. You'll need GnuTLS for the
 >>>>> SSL/ TLS. I haven't modified the mvpmc makefiles to take this into
 >>>>> account, but I assume it's as simple as adding -lgnutls to the
 >>>>> linker flags.
 >>>> Sadly not. The mvpmc build system also builds all the required
 >>>> libraries that are installed on the embedded MVP device. At the
 >>>> moment i'm not aware of anything on the mvpmc that requires GnuTLS,
 >>>> so someone will have to setup the mvpmc build system to build the
 >>>> GnuTLS library aswell.
 >>> That's unfortunate. Just FYI, gnutls requires libgcrypt, which
 >>> requires libgpg-error. However, all three libs are easily cross
 >>> compiled in my experience. I was able to incorporate them into my
 >>> iPhone project in Xcode in a matter of about an hour or two.
 >
 >

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Tom Metro-9]

Eric Lund wrote:
> May I suggest ifdefs here?

I agree.


Chase Douglas wrote:
> I would however suggest that either all of my changes in relation to
> MTP are ifdef'd or none are.

That sounds a bit like throwing the baby out with the bathwater...

MTP sounds like it would be of great use to mvpmc users, almost of none
of which would require the added security of encryption.


> Basically, MTP requires TLS-SRP for any of the added functionality. I  
> don't want anyone enabling MTP and using it in an unencrypted fashion,  
> exposing the completely unsecure MythTV protocol to the greater  
> Internet.

Is there anything about MTP that intrinsically links it to being used on
the Internet? Certainly the vast majority of Mythtv clients are used
within a private LAN, and the MVP platform is just one obvious
application for a transcoded stream that wouldn't typically be used
across the Internet. I hear it is also becoming more common for people
to use low res WiFi connected devices, like Nokia N810 net tablets, to
view video content within their homes.

So I'd say there is a sizable audience for a version that doesn't
require the overhead of an encrypted connection.

Of course this means the ifdefs will need to be added to the MTP server,
as well as the modified client.

  -Tom


------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[MVallevand]

On Wed, Feb 18, 2009 at 10:24 AM, Simon Hyde <simon@...> wrote:

> FWIW this all seems a little over-complicated (as does Roger Heflin's
> transcode script). I find the command line required for ffmpeg to be much
> simpler than for mencoder.

For the record I only suggested mencoder to extract the ac3 track,
ffmpeg is my preferred transcoder.   Unfortunately I have received 50
or more different ts files from various places where vcodec copy and
acopy copy don't work and this has worked better for sync for SD on
the MVP in most cases

mencoder infile -oac copy -ovc copy -of mpeg -mpegopts format=dvd:tsaf
-o outfile

This can be debated forever, but I am quite sure there is no
one-size-fits all solution.

Martin

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Chase Douglas-4]

On Feb 18, 2009, at 6:09 PM, Tom Metro wrote:

> Eric Lund wrote:
>> May I suggest ifdefs here?
>
> I agree.
>
>
> Chase Douglas wrote:
>> I would however suggest that either all of my changes in relation to
>> MTP are ifdef'd or none are.
>
> That sounds a bit like throwing the baby out with the bathwater...
>
> MTP sounds like it would be of great use to mvpmc users, almost of  
> none of which would require the added security of encryption.
>
>
>> Basically, MTP requires TLS-SRP for any of the added functionality.  
>> I  don't want anyone enabling MTP and using it in an unencrypted  
>> fashion,  exposing the completely unsecure MythTV protocol to the  
>> greater  Internet.
>
> Is there anything about MTP that intrinsically links it to being  
> used on the Internet? Certainly the vast majority of Mythtv clients  
> are used within a private LAN, and the MVP platform is just one  
> obvious application for a transcoded stream that wouldn't typically  
> be used across the Internet. I hear it is also becoming more common  
> for people to use low res WiFi connected devices, like Nokia N810  
> net tablets, to view video content within their homes.
>
> So I'd say there is a sizable audience for a version that doesn't  
> require the overhead of an encrypted connection.
>
> Of course this means the ifdefs will need to be added to the MTP  
> server, as well as the modified client.

I understand your point here. However, there's very little reason not  
to use encryption in my opinion. Keep in mind that encryption is not  
enforced for data connections to MTP, only event and control  
connections. That means there's very little compute overhead involved.  
Secondly, GnuTLS is based on LGPLv2, so there's no real license issues  
with respect to libcmyth and mvpmc. Lastly, GnuTLS seems to be very  
portable and well maintained.

Given that there is not much drawback to using GnuTLS and encrypting  
the session, I would be very hesitant to add support for unencrypted  
MTP usage. I really believe that no connection where user data can be  
queried and/or server state can be modified should occur in the clear  
unless both:

1. There's a clear implementation issue technically, licensing wise,  
or otherwise
2. Users completely understand that the service is limited to  
completely trusted networks

Here, I don't believe 1. holds up, and the distinction of 2. is  
blurred by the fact that MTP is partly aimed at use across unsecured  
networks. If someone wants to show how 1. isn't valid, I'll consider  
changing MTP myself. Otherwise, I'd prefer if other people rolled  
their own sources/patches for unencrypted communication.

Thanks

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Eric Lund]

Chase,

The main objection to adding any overhead to libcmyth (and the reason I
suggested ifdefs) is that the MediaMVP platform, the original platform
for mvpmc, is perennially memory and compute starved.  Any additional
resources consumed by mvpmc come as a trade-off against other
functionality.  So, I agree with Tom that, if MTP is useful in a
self-contained, trusted network without an added security layer, it
might be useful to ifdef only the parts that bring in extra dependencies
and leave the rest as functional as possible.

I will say, though, that for the purpose of preserving your work, which
has value in itself, ifdef'ing around the whole thing would suffice.

                                       Eric

Chase Douglas wrote:
 > On Feb 18, 2009, at 6:09 PM, Tom Metro wrote:
 >> Eric Lund wrote:
 >>> May I suggest ifdefs here?
 >> I agree.
 >>
 >>
 >> Chase Douglas wrote:
 >>> I would however suggest that either all of my changes in relation to
 >>> MTP are ifdef'd or none are.
 >> That sounds a bit like throwing the baby out with the bathwater...
 >>
 >> MTP sounds like it would be of great use to mvpmc users, almost of
 >> none of which would require the added security of encryption.
 >>
 >>
 >>> Basically, MTP requires TLS-SRP for any of the added functionality.
 >>> I  don't want anyone enabling MTP and using it in an unencrypted
 >>> fashion,  exposing the completely unsecure MythTV protocol to the
 >>> greater  Internet.
 >> Is there anything about MTP that intrinsically links it to being
 >> used on the Internet? Certainly the vast majority of Mythtv clients
 >> are used within a private LAN, and the MVP platform is just one
 >> obvious application for a transcoded stream that wouldn't typically
 >> be used across the Internet. I hear it is also becoming more common
 >> for people to use low res WiFi connected devices, like Nokia N810
 >> net tablets, to view video content within their homes.
 >>
 >> So I'd say there is a sizable audience for a version that doesn't
 >> require the overhead of an encrypted connection.
 >>
 >> Of course this means the ifdefs will need to be added to the MTP
 >> server, as well as the modified client.
 >
 > I understand your point here. However, there's very little reason not
 > to use encryption in my opinion. Keep in mind that encryption is not
 > enforced for data connections to MTP, only event and control
 > connections. That means there's very little compute overhead involved.
 > Secondly, GnuTLS is based on LGPLv2, so there's no real license issues
 > with respect to libcmyth and mvpmc. Lastly, GnuTLS seems to be very
 > portable and well maintained.
 >
 > Given that there is not much drawback to using GnuTLS and encrypting
 > the session, I would be very hesitant to add support for unencrypted
 > MTP usage. I really believe that no connection where user data can be
 > queried and/or server state can be modified should occur in the clear
 > unless both:
 >
 > 1. There's a clear implementation issue technically, licensing wise,
 > or otherwise
 > 2. Users completely understand that the service is limited to
 > completely trusted networks
 >
 > Here, I don't believe 1. holds up, and the distinction of 2. is
 > blurred by the fact that MTP is partly aimed at use across unsecured
 > networks. If someone wants to show how 1. isn't valid, I'll consider
 > changing MTP myself. Otherwise, I'd prefer if other people rolled
 > their own sources/patches for unencrypted communication.
 >
 > Thanks
 >
 >
------------------------------------------------------------------------------
 > Open Source Business Conference (OSBC), March 24-25, 2009, San
Francisco, CA
 > -OSBC tackles the biggest issue in open source: Open Sourcing the
Enterprise
 > -Strategies to boost innovation and cut costs with open source
participation
 > -Receive a $600 discount off the registration fee with the source
code: SFAD
 > http://p.sf.net/sfu/XcvMzF8H
 > _______________________________________________
 > Mvpmc-devel mailing list
 > Mvpmc-devel@...
 > https://lists.sourceforge.net/lists/listinfo/mvpmc-devel
 >

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel

[Chase Douglas-4]

On Feb 18, 2009, at 9:07 PM, Eric Lund wrote:

> Chase,
>
> The main objection to adding any overhead to libcmyth (and the  
> reason I
> suggested ifdefs) is that the MediaMVP platform, the original platform
> for mvpmc, is perennially memory and compute starved.  Any additional
> resources consumed by mvpmc come as a trade-off against other
> functionality.  So, I agree with Tom that, if MTP is useful in a
> self-contained, trusted network without an added security layer, it
> might be useful to ifdef only the parts that bring in extra  
> dependencies
> and leave the rest as functional as possible.

To give you an idea, the iPhone environment has much the same  
limitations. The api to access the hw decoder on the device does not  
allow true video streaming, so I have to write my own video player  
using ffmpeg. I've only been able to get usable framerates by setting  
the video fps to 10, video bit rate to 200 kbps, 256x256 resolutions  
(so we don't have to scale before rendering to a texture), and 64 kbps  
mono audio (I may drop this down a tad). This nearly pegs the  
processor (and does so when the iPhone tries to do other things in the  
background like checking mail) by itself. I was worried about overhead  
myself, but when I enabled GnuTLS I didn't see any noticeable  
performance decrease, except in the connection handshaking phase.  
GnuTLS does seem to cause connection creation to take twice as long  
(this is how it feels, I haven't measured anything). I also doubt that  
it is eating much memory as well. On the iPhone, you're not guaranteed  
a set amount of memory for your app, but I can say from experience  
that if you go over 30 MB you are likely to get killed. I don't think  
I'm anywhere near there with my app.

> I will say, though, that for the purpose of preserving your work,  
> which
> has value in itself, ifdef'ing around the whole thing would suffice.

If you don't plan on using it for mvpmc, there's no harm in not  
pulling my tree. I still intend to keep my tree as I develop libcmyth  
for my uses and even after it becomes stable. If you pull all my work,  
then that's definitely nice, but it's not necessary.

In fact, I wouldn't really pull my tree yet unless you are interested  
in testing/developing it as I know of some areas of MTP that aren't  
complete yet. It's certainly not in a state that should be used for  
production level programs. I certainly welcome any development help I  
can get though.

Thanks

------------------------------------------------------------------------------
Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco, CA
-OSBC tackles the biggest issue in open source: Open Sourcing the Enterprise
-Strategies to boost innovation and cut costs with open source participation
-Receive a $600 discount off the registration fee with the source code: SFAD
http://p.sf.net/sfu/XcvMzF8H
_______________________________________________
Mvpmc-devel mailing list
Mvpmc-devel@...
https://lists.sourceforge.net/lists/listinfo/mvpmc-devel